ARTICLE

https://doi.org/10.1038/s41467-021-21120-8 OPEN Mechanism of the small ATP-independent chaperone Spy is substrate specific

Rishav Mitra1,2, Varun V. Gadkari3, Ben A. Meinen1,2, Carlo P. M. van Mierlo 4, Brandon T. Ruotolo 3 & ✉ James C. A. Bardwell1,2

ATP-independent chaperones are usually considered to be holdases that rapidly bind to non- native states of substrate proteins and prevent their aggregation. These chaperones are

1234567890():,; thought to release their substrate proteins prior to their folding. Spy is an ATP-independent chaperone that acts as an aggregation inhibiting holdase but does so by allowing its substrate proteins to fold while they remain continuously chaperone bound, thus acting as a foldase as well. The attributes that allow such dual chaperoning behavior are unclear. Here, we used the topologically complex protein apoflavodoxin to show that the outcome of Spy’s action is substrate specific and depends on its relative affinity for different folding states. Tighter binding of Spy to partially unfolded states of apoflavodoxin limits the possibility of folding while bound, converting Spy to a holdase chaperone. Our results highlight the central role of the substrate in determining the mechanism of chaperone action.

1 Howard Hughes Medical Institute, University of Michigan, Ann Arbor, MI, USA. 2 Department of Molecular, Cellular, and Developmental Biology, University of Michigan, Ann Arbor, MI, USA. 3 Department of Chemistry, University of Michigan, Ann Arbor, MI, USA. 4 Laboratory of Biochemistry, Wageningen ✉ University, Wageningen, The Netherlands. email: [email protected]

NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8

opologically complex proteins often populate misfolded bound to the chaperone10. The substrates studied in detail so far, intermediates that act as kinetic traps1.Suchinter- namely Im7 and SH3, are small proteins (10 and 7 kDa, respec- T α β mediates often expose hydrophobic surfaces that make tively) with very simple 3D structures (just -helices or just - them prone to aggregation. ATP-dependent molecular cha- strands, respectively). We wondered if the folding-while-bound perones like GroEL-GroES rescue trapped intermediates and mechanism of Spy applies to more complex substrates that include facilitate substrate folding2. In contrast, ATP-independent both α-helices and β-strands and have a more complex folding chaperones generally bind very tightly to non-native sub- pathway than the substrate proteins tested so far. To investigate this, strates and in doing so prevent protein aggregation but are not we chose the apoflavodoxin AnFld as our substrate, as its native thought to directly facilitate substrate refolding3.Thesimple state has an α–β parallel topology and as apoflavodoxin exhibits a designations of “foldase” or “holdase” may underemphasize the more complex folding pathway than either Im7 or SH3 in that it microscopic structural heterogeneity of chaperone–substrate folds via a three-state triangular mechanism with an essentially off- complexes4. Substrates bound to ATP-independent chaperones pathway intermediate (Fig. 1a)11. In this mechanism, most of the such as trigger factor, SecB, and the sHsps can adopt a wide molecules in the intermediate conformation need to unfold prior to range of conformations ranging from near-native to unfolded refolding into the native conformation. A small fraction of mole- states5–7. We have shown that the ATP-independent chaperone cules can directly fold from the intermediate to the native state. Spy not only binds protein folding intermediates of substrate First, we verified that the previously established triangular folding proteins such as Im7 and SH3 but also allows for the folding of mechanism for flavodoxin folding is functional under our buffer substrate proteins while they remain chaperone bound8,9.Spy conditions (40 mM HEPES-KOH pH 7.5, 100 mM NaCl), which we loosely binds to different folding states of these model substrate considered to be more physiological than the previously used buffer proteins.Althoughthefoldingrateconstantforthesetwo (50 mM MOPS pH 7.0) (Fig. S1a–d)8,11. We studied the refolding simple substrates decreases with increasing concentrations of of AnFld in the presence of Spy by monitoring the tryptophan Spy, it does not go to zero at saturating Spy concentrations fluorescence in a stopped-flow fluorimeter upon diluting urea- where essentially all the substrate molecules are chaperone denatured AnFld into refolding buffer containing increasing bound10, evidence that folding of the substrate occurs while it concentrations of Spy, as described previously8. We observe two chaperone-bound. In this mechanism, substrate release is not a kinetic phases of AnFld folding; a major kinetic phase, where the prerequisite for substrate folding. We have found that the amplitude is proportional to the fraction of denatured molecules folding-while-bound mechanism is dependent on relatively that directly fold to the native state (N), and a minor phase whose weak chaperone–substrate interactions9.VariantsofSpythat amplitude is proportional to the fraction of molecules that bind SH3 with stronger affinity than does wild-type (WT) Spy transiently populate the off-pathway intermediate and therefore significantly slow the folding of bound substrate9.Thesub- need to unfold before productive refolding11. In the presence of Spy, strates of Spy tested so far have been small, topologically sim- the observed rate constants (kobs) of both phases decrease ple, all β-sheet or all α-helical model proteins. We wanted to substantially until at the highest Spy concentration (~9 μM), a −1 test if the folding-while-bound model is generalizable to larger single folding phase remains that has a small kobs value (0.1 s ) substrates with complex topologies such as those found in the (Fig. 1b, c). This indicates that Spy significantly slows the folding vast majority of proteins. Here, we chose apoflavodoxin from rate of AnFld, like for Im7 and SH3. However, both Im7 and SH3 Anabaena PCC 7119 (AnFld) and Azotobacter vinelandii can fold to the native state, albeit slowly, even at saturating Spy (AzoFld), both well-studied folding models11,12 that populate concentrations8,9. In contrast, binding to Spy prevents folding of kinetically trapped off-pathway intermediates13,14.Theirα/β AnFld almost completely, as evidenced by the dramatic decrease in topology is both ancient and frequently found in common folds the folding amplitude upon increasing Spy concentration (Fig. 1b). such as TIM barrels, Rossman folds, FAD/NAD(P)-binding At Spy concentrations higher than ~9 μM, the amplitude of the domains, and P-loop containing hydrolases. The goal of this fluorescence change was extremely small, indicating that most of study was to test whether the folding-while-bound mechanism the denatured AnFld molecules are kinetically trapped in a folding- adequately describes the effect of Spy on the folding of proteins incompetent bound state. Both the unfolded (U) and intermediate with complex topologies using the flavodoxin-like fold as an (I) states of AnFld are less fluorescent than the native state11.Since example. Our results show that Spy’s mechanism is surprisingly AnFld in the presence of Spy is substantially less fluorescent than it substrate specific.Inthecaseofapoflavodoxin, Spy traps is in the native state, we reason that Spy kinetically traps AnFld in a denatured polypeptides in a non-native state, thereby inhibiting non-native state that is partially or completely unfolded (we term flavodoxin folding. This dual functionality of Spy as both a this state U*). It appears that Spy binds very rapidly to AnFld and folding-while-bound chaperone and a holdase rests on oppos- prevents folding since the fluorescence decreases within the ~25 ms ing thermodynamic requirements. The folding-while-bound dead time of the stopped-flow instrument. The interaction of Spy * paradigm requires that Spy bind the various folding states of and the U state is tight with a dissociation constant (KD)of0.35 the substrate weakly. In contrast, a holdase binds non-native μM(Fig.1d). The affinity of Spy for AnFldU* in folding states tightly such that folding transitions are not feasibly experiments is ~13-fold tighter than observed for Spy binding to = μ thermodynamically. We show that the mechanism of action of the intermediate and unfolded states of Im7 (KD 4.7 M for both) the chaperone Spy is dictated by the difference in binding and ~8-fold tighter than that between Spy and unfolded SH3 fi = μ 8,9 af nities for partially unfolded states of different substrates to (KD 2.9 M), both of which fold while bound to Spy .The Spy. This work highlights the need to shift from monolithic kinetic trapping of AnFld in the Spy-bound U* state due to these enzyme-inspired views of molecular chaperones toward more high-affinity interactions apparently necessitates release from Spy complex models that incorporate the possibility of multiple before productive folding can occur. substrate-specific mechanisms.

Rapid binding of Spy is coupled to partial unfolding of Results apoflavodoxin. Having shown that Spy inhibits the folding of Spy inhibits the folding of apoflavodoxin by kinetically trap- AnFld, we wondered if Spy interacts with AnFld’s native state. ping it in a non-native state. Spy belongs to an emerging class of Holdase chaperones are known to preferentially bind to non- chaperones that allow protein folding while the substrate remains native unfolded or misfolded proteins15. Folding-while-bound

2 NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8 ARTICLE

Fig. 1 Kinetics of apoflavodoxin folding in the presence of Spy. a Folding pathway of apoflavodoxin from Anabaena PCC 7119 (AnFld) and Azotobacter 11, 12 vinelandii (AzoFld) . N, U, and I are the native, unfolded, and intermediate states, respectively. In case of AzoFld, Ioff and Ion are the off-pathway and on- pathway intermediates, respectively. b Fluorescence traces for the refolding of 0.09 µM AnFld in the presence of various concentrations of Spy dimer (0–9.1 μM after mixing). Time axis is shown in logarithmic scale. AU arbitrary units. The traces of AnFld refolding in the presence of 0–0.5 μM Spy dimer could be fit to a sum of two exponentials, and traces for higher Spy concentrations could be adequately fit to a single exponential function. The absence of the fast phase at higher Spy concentrations most likely indicates that Spy completely blocks the direct folding pathway from unfolded to native AnFld. c Plot of the observed rate constants (kobs) for AnFld WT folding as a function of Spy dimer concentration. d Initial fluorescence values were calculated from the fit of the kinetic datasets. AU arbitrary units. The binding isotherm was plotted with normalized initial fluorescence intensity during AnFld WT refolding in the y-axis and fitted to a one-site binding model. All experiments were performed in HN buffer at 25 °C. Value reported is the mean ± s.e.m. of the fit. Source data are provided as a Source Data file. chaperones, on the other hand, also bind to the native state. binding affinities for diverse substrates indicates that weak Thus, one way to distinguish between holdases and foldases is (micromolar) affinity for native substrates is apparently a to determine the relative affinities of the chaperone for native general property of Spy. This makes sense; tight binding of a and non-native states. If Spy is functioning as a holdase with chaperone to the native state might well inhibit substrate AnFld we expect Spy to bind the native state with much weaker function. Whether Spy inhibits folding or allows the substrate affinity than fully and partially unfolded states. We found a to fold while bound may therefore depend mainly on the affi- binding stoichiometry of 1:1 between AnFld and the Spy dimer nity that Spy has for non-native states. using analytical ultracentrifugation (AUC) (Fig. S2a)8,9.There We observed a slow exponential decrease in fluorescence of −1 is a very rapid decrease in fluorescence upon mixing Spy and AnFld in the presence of Spy (kobs = 0.04 s ) after the burst phase AnFld (Fig. 2a). The amplitude of this burst phase reaches (Fig. 2b). This decrease is not due to Spy quenching AnFld saturation at high Spy concentrations, consistent with very fluorescence because the final fluorescence decreases hyperbolically rapid binding of Spy to the native AnFld (Fig. S2b). However, with Spy concentration (Fig. S2c). Instead, we suggest that Spy the interaction is weak, with an apparent dissociation constant binding leads to a conformational change in AnFld. This (KDapp) of 40 µM, 114-fold weaker than the KD for Spy with conformational change may involve Spy-mediated unfolding of a U* AnFld (Fig. 2b). The KDappfor native AnFld is similar to that minor population of native AnFld to a thermal intermediate-like previously observed for the interaction of Spy with native Im7 state (IT), which was observed to be less fluorescent than the native = = 8,9 16 T (KD 20.5 µM) and SH3 (KD 50 µM) . These comparable state .ThenativeandI states have largely similar structures that

NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8

(a) (b) Burst phase

[Spy dimer] (M) 0.0 13.7 27.4 45.7 54.8 K  D,app = 39.9 ± 1.1 M

1.8 (c) 0.35 (d) 0.4

0.3 1.8 1.6 Partial Concentration Partial Concentration 0.3 0 0.25 0 1.6

1.4 0.2 1.4 0.2

0.15 1.2 Frictional Ratio f/f Frictional Ratio f/f 1.2 0.1 0.1

1 1 0.05

0.0 0 1.5 2 2.5 3 3.5 1.5 2 2.5 3 Sedimentation Coefficient (1e-13) in water at 20°C Sedimentation Coefficient (1e-13) in water at 20°C (e) (f) 0.05 0.0 0.00

-0.05 -1 N = 0.83 ± 0.02 sites -0.10 -1 of injectant -2.0 K = 8.01e5 ± 2.20e5 M -1 ΔH = -4087 ± 131.8 kcal mol-1 cal sec -1 -1  -0.15 ΔS = 12.6 cal mol deg

-0.20 kcal mol

-0.25 -4.0

0 50 100150200 0.0 0.5 1.0 1.5 2.0 2.5 Time (min) Molar Ratio of Spy dimer

Fig. 2 Interaction of Spy and native apoflavodoxin. a Change in intrinsic tryptophan fluorescence of 0.09 µM AnFld WT with time in the absence and presence of various concentrations of Spy (0–54.8 μM after mixing). Time axis is shown in logarithmic scale. AU arbitrary units. The black lines shown for each Spy concentration are best fits of each dataset to a single exponential equation. b AnFld WT-Spy binding as monitored by intrinsic tryptophan fluorescence of AnFld. Spy dimer was titrated into 17.2 µM native AnFld WT at 25 °C. The binding isotherm is plotted with change in fluorescence (F)as normalized by fluorescence intensity relative to native AnFld fluorescence (F0) in the y-axis. Black line shows the fit to a one-site binding model. c, d SV- AUC experiments showing frictional ratios (f/f0) and sedimentation coefficients for c AnFld WT alone and d AnFld WT in the presence of 2.5-fold excess of Spy dimer in 40 mM HEPES-KOH (pH 7.5), 25 mM NaCl. Data were analyzed by two-dimensional sedimentation analysis (2DSA) followed by analysis with a genetic algorithm, which was further validated by a Monte Carlo analysis. e, f ITC analysis for the interaction of Spy and AnFld F98N at 10 °C. The low temperature minimizes enthalpy changes due to the binding-induced unfolding of AnFld. Five hundred and fifty micromolar Spy dimer in the syringe was titrated into 50 μM AnFld F98N in the cell. The thermogram in e was integrated and fit to a one-site binding model, as shown in f to obtain thermodynamic binding parameters. N stoichiometry of binding, K association constant, ΔH enthalpy change, ΔS entropy change. Values reported are the mean ± s.e.m. of the fit. Source data are provided as a Source Data file. include two hydrophobic cores formed by α-helices packing onto a concentrations of Spy, the fluorescence of the AnFld WT–Spy central β-sheet. Only the loop connecting β4andα4andthelong complex at 340 nm is 0.33 times that of the fluorescence of loop splitting the strand β5 are disordered in IT. To further explore AnFld WT alone (Fig. 2b), a value very close to the fluorescence of the nature of bound AnFld, we used a previously characterized AnFld F98N (0.39 times that of the WT) (Fig. S2d). This similarity mutant, F98N, that mimics the IT state17. At saturating in fluorescence values supports the idea that binding of Spy is

4 NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8 ARTICLE coupled to partial unfolding of AnFld to an IT-like state. Additional Spy forms a compact chaperone–substrate complex with con- support for this proposed partial unfolding comes from studying formationally heterogeneous apoflavodoxin. To probe the the interaction of Spy and AnFld in low ionic strength solutions. interaction between Spy and AnFld, we employed native mass AUC experiments revealed that in HEPES buffer containing very spectrometry (MS) which is capable of detecting non-covalent low salt concentrations (25 mM NaCl), AnFld WT also exists in protein–protein interactions in the gas phase. Native MS is predominately two conformations that have the same sedimenta- particularly well-suited for the analysis of mixtures such as co- tion coefficient but different frictional ratios (f/f0)(Fig.S2eand incubated proteins and their complexes, as each population is Fig. 2c). The major conformation (native state) contributes ~60% of separated by mass20. When coupled to ion mobility (IM), native the absorbance signal and has a frictional ratio close to one, typical IM-MS enables gas phase structural measurements of the mass- of globular proteins. The minor conformation contributes ~40% of resolved proteins and complexes. The IM arrival time distribu- the signal resembles the partially unfolded IT in that it is more tions (ATD) of individual ions can be converted to rotationally expanded with a frictional ratio of ~1.7. In the presence of excess averaged collision cross-section distributions (CCSD). The cen- amounts of Spy, only a single AnFld WT species is seen (Fig. S2e troid of these distributions can be reported as the rotationally DT and Fig. 2d) indicating that the conformational heterogeneity of averaged collision cross section ( CCSN2, CCS). CCS is a charge AnFld seen under non-denaturing conditions is lost upon independent structural parameter that correlates with the surface interaction with Spy. We cannot ascertain whether in the low area of the ion21. It is sensitive to domain-level rearrangements ionic strength buffer, only the IT state of AnFld binds Spy and in that occur in solution prior to ionization. For instance, a rigid doing so pulls the equilibrium away from the native state, or protein with a single solution conformation exhibits a Gaussian whether Spy also binds native AnFld and partially unfolds it to a ATD/CCSD, whereas a dynamic protein exhibits a multimodal or state that resembles IT. Under physiological salt concentrations extended ATD/CCSD22. however, most of the AnFld molecules populate the native state Using native IM-MS, we first compared WT AnFld and the (Fig. S2f). Isothermal titration calorimetry (ITC) experiments destabilized 2A mutant. The CCSs of AnFld 2A are similar to showed that Spy binds the WT and F98N mutant with dissociation those of AnFld WT, indicating that both sample a similar constants of 28.2 and 1.8 µM, respectively (Fig. S2g, S2h and Fig. 2e, structural ensemble (Fig. S4b and Table S2). The small f). The ~16-fold higher affinity of Spy for the IT mimic than for WT difference observed in CCS can be attributed to the small shift should shift the equilibrium of Spy-bound AnFld towards the more in mass of 2A vs. WT. The broad CCSD of both AnFld’s(Fig. expanded thermal intermediate. Thus, the molecular basis for 4 and Fig. S4b), ranging from 18 to 20 nm2 for the compact binding-induced partial unfolding of native substrates appears to be states, and 20 to 45 nm2 for the extended unfolded states, indi- the higher affinity of Spy to more unfolded conformations8. cates that the proteins are structurally heterogeneous in solution. The highest charge states exhibit collision cross- sections equal to those observed for proteins nearly two- to Spy binds tightly to a partially unfolded mutant of apo- three-fold the size of apoflavodoxin23,56. Such conformational flavodoxin. If the binding of Spy tightly to partially or fully heterogeneity is expected as AnFld is destabilized in low ionic unfolded conformations of AnFld inhibits AnFld refolding, strength buffers like the 20 mM ammonium acetate solution then Spy should have a higher affinity to destabilized mutants used in these experiments16,24. The use of this buffer is of AnFld than it does to the WT protein. We combined two fortuitous as it gives us an opportunity to detect the native state destabilizing mutations in AnFld (L105A and I109A)18 to and otherwise sparsely populated partially folded states of generate AnFld 2A. Sedimentation equilibrium experiments AnFld. The degree of protein folding can also be assessed by showed that this protein is monomeric and soluble (Fig. S3a). inspection of the charge state distribution (CSD) of ions. Multiple lines of evidence indicate AnFld 2A is partially Natively folded compact proteins with rigid structures have a unfolded. CD spectra indicate the A2 mutant has lost about CSD spanning 2–4 charge states. In contrast, both AnFld 30% of its alpha helical content (Fig. 3aandTableS1),though proteins analyzed in this work have a broad multimodal CSD it β-sheet content is unchanged (Table S1). Other evidence spanning 14 charge states, including “completely unfolded” that the 2A mutant is partially unfolded include a decrease in higher charge states (12+ to 19+), some “partially unfolded” intensity and a redshift in tryptophan fluorescence (Fig. 3b) charge states (9+ to 11+), and some “folded”, compact charge and an increase in ANS (1-anilino-8-napthalene sulfonic acid) states (6+ to 8+)(Fig.S4)confirming that these proteins binding (Fig. 3c). Sedimentation profiles show that the mutant populate various conformations in solution. The unfolded 9+ is expanded relative to the WT protein (Fig. S3a) and urea to 19+ charge states account for 15 ± 2% of total AnFld 2A denaturation profiles indicated that the mutant is highly signal intensity, while only accounting for 1.5 ± 0.6% of total destabilized (Fig. 3d) and it unfolds more rapidly than WT AnFld WT signal intensity (Fig. S4a). Although AnFld WT and (Fig. S3b). Limited dispersion in the 1H dimension of the 2D AnFld 2A both apparently sample the same structural [1H−15N] HSQC–TROSY NMR spectra of the mutant ensemble, the equilibrium of AnFld 2A is clearly more shifted between 7.5 and 8.5 ppm indicates that it is mostly disordered, towards the unfolded states compared to AnFld WT. though 48 residues maintain native-like backbone 15Nshifts Next, we analyzed complexes formed between Spy, and both (Fig. 3e and Fig. S3c, S3d)19. Taken together, our experiments AnFld variants. Three CSDs were observed between 2000 and indicate that the AnFld 2A mutant is partially unfolded. 4000 m/z corresponding to AnFld, Spy and Spy in complex with Fluorescence experiments indicate that Spy rapidly binds the AnFld (Fig. 4a, b and Table S2). The Spy:AnFld complexes, 2A mutant with a KD app of 0.1 µM (Fig. 3e and S3e). ITC and containing either AnFld WT or AnFld 2A, consistently exhibit kinetic unfolding experiments indicate that AnFld 2A exists in lower CCSs than the larger more unfolded forms of AnFld (z = two binding-competent conformations that interact with Spy with 15+ to 19+). comparable affinities of 0.38 and 0.23 µM (Fig. S3f and S3b). Since Spy does not undergo major changes in structure and These affinities measured for AnFld 2A are very similar to the 0.4 dynamics upon substrate binding25,26, we attribute the changes μ * MKD found for Spy interaction with the U state of AnFld WT in CCS upon complex formation to changes in the conforma- (Fig. 3f). Collectively these observations support our model that tion of AnFld upon Spy binding. Measurements of intermole- Spy kinetically traps AnFld in a partially unfolded state by rapid cular paramagnetic relaxation enhancement effects in the and tight binding, thus functioning as a holdase. Spy–Im7 complex by NMR spectroscopy showed that binding

NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8

(a) (b) AnFld WT AnFld WTN AnFld 2A AnFld 2A AnFld WTU

(c) AnFld WT (d) AnFld WT AnFld 2A AnFld 2A Buffer

(e) AnFld WT (f)

11.0 0.5 10.0 9.5 9.0 8.5 8.0 7.5 7.0 6.5  K D,app = 0.1 ± 0.06 M

AnFld 2A

11.0 0.5 10.0 9.5 9.0 8.5 8.0 7.5 7.0 6.5 ¹H (ppm)

Fig. 3 Characterization of the AnFld2A mutant and its interaction with Spy. a Far-UV CD spectra of AnFld WT (red) and AnFld2A mutant (green) in 0.1 M potassium phosphate buffer (pH 7.5) at 25 °C. b Intrinsic tryptophan fluorescence emission spectra of the native states of AnFld WT (green), AnFld 2A mutant (blue), and denatured AnFld WT in 5 M urea (red) in HN buffer at 25 °C. c Fluorescence emission spectra of ANS alone (red) in HN buffer and in the presence of AnFld WT (green) or AnFld 2A mutant (blue) in HN buffer at 25 °C. The fluorescence intensity of ANS increases 2.8-fold upon binding the mutant compared to the WT. d Urea-induced denaturation of AnFld WT (red) and AnFld 2A mutant (blue) monitored by tryptophan fluorescence emission at 340 nm in HN buffer at 25 °C. The data were normalized to the fluorescence signal (F340) of each protein in the absence of urea. The denaturation curve of the WT protein was fit to a two-state unfolding model (solid black line) and that of AnFld 2A was fitted to a third-degree polynomial (broken black line). e 1H dimension of the 2D [1H-15N] HSQC–TROSY spectra of 0.2 mM AnFld WT (red) and 2 A (blue). f Titration of Spy dimer to 10 µM AnFld 2A as monitored by the intrinsic tryptophan fluorescence. The black line shows the fit of the fluorescence intensity as a function of Spy dimer concentration to a tight binding model. A.U. arbitrary units. Value reported is the mean ± s.e.m. of the fit. Source data are provided as a Source Data file. of Spy induces compaction of a dynamic ensemble of unfolded results from spectroscopy-based kinetic experiments and IM- Im7 L18A L19A L37A26.Thecompactionofanunfoldedclient MS reveal that such client compaction may also stabilize in complex with Spy is proposed to facilitate intramolecular kinetically trapped states that are unable to fold in a chaperone- contacts necessary to fold while bound26,27. Collectively, our bound form.

6 NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8 ARTICLE

50 AnFld WT SpyD:AnFld WT (a) (b) 45 AnFld WT SpyD SpyD:AnFld 2A SpyD 6+ 19+ D 18+ Spy -AnFld WT 17+ 7+ ) 40 40 10+ 2 16+ 16+ 15+ 15+ 14+ 13+

(nm 14+ 8+ 35 11+ N2 13+ 30 9+ CCS 13+

DT 30 13+ 12+ Drift Time (ms) 12+ 11+ 10+ 14+ 15+ 20 11+ 25 16+ 10+ 9+ 12+ 8+ 7+ 6+

1000 1500 2000 2500 3000 3500 4000 20002500 3000 3500 4000 m/z m/z

Fig. 4 Native IM-MS to study AnFld and its complex with Spy. a Collision cross-sections of AnFld WT (orange circles), Spy (red circles), Spy:AnFld WT (blue circles) versus m/z. The complex of Spy:AnFld 2A (blue triangles) is also plotted for comparison of the complexes. b Ion mobility mass spectrum of AnFld WT (orange) co-incubated with SpyD (red). Complexes of SpyD:AnFld WT are highlighted in blue.

Chaperone-binding surface of apoflavodoxin is composed of binding to partially unfolded states. Given the striking differences dynamic regions implicated in folding. Using NMR spectro- between Spy’s mechanism of action for AnFld and simple substrates scopy, we set out to characterize the binding surface of AnFld at like Im7 and Fyn SH3, we wished to test the holdase activity of Spy the residue level. Upon the addition of sub-stoichiometric with apoflavodoxin from another species, Azotobacter vinelandii, amounts of unlabeled Spy, we observed a loss of signal inten- that populates a stable molten globule intermediate12.Moltenglo- sity, as quantified by the peak heights in the (2D) [15N, bules are a class of compact folding intermediates that lack persistent 1H]–TROSY NMR spectra of [U-2H, 15N, 13C]-labeled AnFld native tertiary structure but possess a significant amount of native- (Fig. S5a). Considering both the increase in size upon complex like secondary structure31.Theflavodoxin from Azotobacter vine- formation (size of apo-AnFld is ~19 kDa and that of the AnFld- landii, which we term AzoFld, shares 47% sequence identity, and is Spy complex is ~50 kDa) and the micromolar affinity of the structurally very similar (RMSD of 0.6 Å) to the flavodoxin from interaction, we reasoned that line broadening can be caused by Anabaena with which we did all of the proceeding experiments increased tumbling time and/or increased chemical exchange in (Fig. S6a). AzoFld folds via a complex four-state mechanism that the µs-ms time scale. We attempted to map the binding surface of involves both the formation of a stable off-pathway molten globule AnFld for Spy by plotting the ratio of peak intensities of AnFld intermediate (MGoff)thatactsasakinetictrapandanobligatehigh- 12 residues in the absence and presence of 0.75× unlabeled Spy. energy on-pathway intermediate (Ion)(Fig.1a) .Themajorkinetic Residues were considered to be part of the interaction site if they phase during refolding reports on the rate-limiting step of unfolding underwent greater than 2/3 of the average loss in peak intensity of the molten globule intermediate formed rapidly after dilution upon Spy binding (Fig. 5a). These residues largely colocalize on from denaturant12. one surface of native AnFld (Fig. 5b, c and Fig. S5b). However, we The sedimentation profiles of AzoFld alone and in the cannot eliminate the possibility that there are additional binding presence of Spy showed that the Spy dimer forms a 1:1 complex sites on AnFld as chemical shift assignments were unavailable for with AzoFld (Fig. S6b). By monitoring the refolding of urea- stretches of contiguous residues G9–S17 at the phosphate-binding denatured AzoFld in buffer containing increasing concentra- site and C54–Y70 at the FMN-binding site28. tions of Spy, we observed that Spy decreased both the folding We note that several residues in the two loops, 90–100 and rate and yield of AzoFld (Fig. 6a). Similar to our observations 120–139, that were found to interact with Spy based on their with AnFld, the decreasing fluorescence amplitude with intensity ratios, have previously been found to be unstructured in increasing Spy concentrations indicates that AzoFld cannot the otherwise natively folded thermal intermediate of AnFld undergo efficient folding when bound to Spy. Kinetic refolding (Fig. S5c)29. Binding of Spy to these dynamic regions on AnFld of AzoFld occurs in two phases in the absence of Spy. The may help stabilize the partially disordered AnFld in a high- minor folding phase disappears in the presence of even low affinity complex. This is consistent with our native IM-MS amounts of Spy (>4.6 µM). The kobs forthemajorfoldingphase decreased hyperbolically with increasing concentrations of Spy experiments, which indicate that the AnFld-Spy complex is more − compact than unfolded conformations of unbound AnFld that and asymptotically reached the very low value of 0.09 s 1 at exist in solution. Interestingly, Monte Carlo simulations using a saturating Spy concentrations (Fig. 6b). The simplest explana- nucleation-growth model showed that the folding nucleus of tion for this observation is that Spy binds more tightly to both AnFld comprises the loop 90–100 and the segment 120–160 that MGoff and U states than it does to native AzoFld, thereby includes α4, β5, most of the C-terminal helix α5 and connecting allowing Spy to inhibit the folding of AzoFld. The small non- loops, in other words almost completely overlaps with the zero kobs value suggests that unfolding of MGoff can still occur binding surface for Spy30. While the implications of this finding while bound to Spy, albeit at a much slower rate. It is are unclear, we speculate that Spy may kinetically trap refolding noteworthy that the U and MGoff states are indistinguishable AnFld in the non-native U* state by tightly binding to a folding species in the absence of denaturant due to a gradual second- nucleus that is formed early during folding. order-like collapse transition of the unfolded protein to the 32 MGoff state under equilibrium conditions . Given that the MGoff state is reported to be aggregation-prone in the presence Spy kinetically traps the misfolded molten globule of another of cytosol-mimicking crowding agents, binding to such apoflavodoxin. We have shown that Spy’s ability to inhibit the misfolded intermediates is additional strong evidence for Spy fl 33,34 folding of Anabaena flavodoxin apparently results from its tightly having a holdase activity for avodoxin-fold substrates .

NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8

(a) 1.0

0.8

0.6 (-Spy) /I

0.4 (+Spy) I 0.2

0.0 0 20 40 60 80 100 120 140 160 Residue number (b) (c)

Fig. 5 NMR spectroscopy to map the Spy binding surface of AnFld. a A plot of the ratios of intensities of each assigned cross peak in the [1H-15N]–TROSY

NMR spectra of AnFld WT in the presence of sub-stoichiometric amount of Spy (0.75×) and in the absence of Spy [I(+Spy) and I(−Spy), respectively]. The asterisks denote unassigned residues in the spectra, and the dashed red line shows the cutoff of average peak broadening for all mapped residues. b, c A red-to-blue color scale has been used to map the NMR peak intensity ratios on the crystal structure of AnFld (Protein Data Bank ID code 1FTG) in b ribbon and c surface representations. Red and blue represent the highest and lowest intensity ratios in the dataset at 0.784 and −0.002, respectively. A lower intensity ratio, i.e., [I(+Spy)/I(−Spy)] is indicative of a higher degree of backbone perturbation in the presence of Spy. The unassigned residues in AnFld are show in gray.

Discussion chaperone Skp for instance, binds the unfolded outer membrane 37 Misfolded and unfolded states of cellular proteins are recognized protein OmpA with a KD of 22 nM . The chaperone activity of by ATP-independent holdase chaperones that act by sequestering the human small heat-shock protein Hsp27 is enhanced upon them to prevent aggregation3. Our results with the small bacterial stress-induced phosphorylation40. The phosphorylation mimic of ATP-independent chaperone Spy showed that simple model Hsp27, S15D/S78D/S82D binds a destabilized T4 lysozyme var- proteins like Im7 and Fyn SH3 can fold while continuously iant with an apparent affinity of 4 nM41. Although the exact bound to Spy, provided the different folding states bind the mechanism varies, these chaperones generally maintain substrates chaperone only weakly10. The relative affinity of Spy to the dif- in an unfolded state, thereby preventing aggregation. That these ferent folding states of the substrate appears to determine whether chaperones bind tightly to non-native states with different substrate folding occurs while bound to Spy or not. We previously degrees of unfolding intuitively supports their holdase activity. In observed that the Spy mutants Q100L and H96L inhibit SH3 the case of Im7 and Spy, the various Im7 folding states that are refolding9. The mechanistic basis for the unfolding activity of bound to Spy can interconvert while chaperone bound apparently these Spy mutants was an increase in the binding affinity of Spy at least in part due to the weaker nature of the interaction25,26.In fi towards unfolded SH3 (KD, unfolded). Furthermore, the af nity of the case of AnFld, the interaction of partially unfolded states with Spy to unfolded SH3 correlates with the folding rate constant in Spy is substantially tighter and this tighter binding apparently the presence of Spy. These Spy mutants are also potent holdase prevents folding of the chaperone-bound substrate. These chaperones that suppress the in vitro aggregation of denatured observations make intuitive sense; tight binding of the chaperone substrates like aldolase and α-lactalbumin35. We postulate that will hinder conformational transitions in the bound substrate and only by binding loosely and with comparable affinity to unfolded, thus prevent folding while bound. The inability of sHsps to work intermediate, and native states, can a chaperone allow folding as folding-while-bound chaperones may be in part due to them while bound without interfering with the substrate protein’s relying on short-range hydrophobic interactions to recognize function. In contrast, binding and sequestration of aggregation- non-native clients39. This mode of client recognition also pre- prone denatured states is a hallmark of ATP-independent holdase vents them from binding native proteins, which normally do not chaperones10. Nanomolar affinities for unfolded polypeptides and expose hydrophobic surfaces. Tight binding to unfolded and folding intermediates have been observed for the interaction of intermediate states renders sHsps incapable of spontaneous substrates with ATP-independent chaperones such as trigger release and refolding of clients. The chaperone function of sHsps factor, Skp, SecB, and sHsps36–39. The periplasmic holdase is therefore limited to sequestering aggregation-prone unfolded

8 NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8 ARTICLE

(a) (b) [Spy dimer] (M) 0.0 4.6 13.7 27.4 45.7 63.9 91.3 118.7 137.0 182.6

(c)

† U U

N N

U-Spy N-Spy N-Spy Energy † U-Spy

Folding-while-bound Holdase

Fig. 6 Spy inhibits the folding of flavodoxin-fold substrates by acting as a holdase. a Fluorescence traces of the refolding of 0.1 μM AzoFld C69A in the presence of various concentrations of Spy dimer (0–182.6 μM after mixing) in KP buffer. Time axis is shown in logarithmic scale. b A plot of the observed major rate constant for AzoFld refolding (kobs) as a function of Spy dimer concentration. c Mechanism of substrate-specific action of Spy. Simple substrates can fold while being continuously chaperone-bound because Spy binds the native (N) and unfolded (U) states with weak affinities. Even at saturating concentrations of Spy, folding occurs because the N–Spy complex is the most energetically stable form of the substrate. However, large and topologically complex substrates cannot fold while bound because of Spy’s strong affinity for (partially) unfolded states (denoted generically by U†). For such substrates, Spy acts as a holdase by sequestering aggregation-prone U† states. Kinetically, such holdase-like action makes Spy a competitive inhibitor of folding because the U†–Spy complex is the most energetically stable form of the substrate. Source data are provided as a Source Data file. proteins38. Spy on the other hand utilizes mainly long-range substrate interaction. Instead, Spy’s interaction with its substrates electrostatic interactions to recognize and bind unfolded sub- is, in some cases (Im7 and SH3), apparently finely tuned by strates27. Although the Spy-unfolded substrate complex is also evolution to allow for loose binding and thus folding while stabilized by hydrophobic interactions, for substrates like Im7 and bound. In other cases, such as the one studied here with apo- SH3, binding is weak and allows the substrate to fold to its native flavodoxin, Spy binds tightly enough to inhibit folding while state27. Our work with apoflavodoxin shows that the affinity of bound thereby mimicking the action of well-studied holdase Spy to partially unfolded substrates is crucial in determining chaperones. For topologically complex model substrates like whether Spy’s chaperone function mirrors that of sHsps (Fig. 6c). apoflavodoxin, binding tighter to aggregation-prone misfolded The substrate-specific mode of action of Spy may enable it to states may be the only productive mechanistic possibility in the be very effective as an ATP-independent chaperone. Spy has a absence of ATPase activity. Collectively, our results reveal a dual function; it can facilitate protein folding by either allowing substrate-specific mechanism for Spy where this chaperone exists folding while bound or it can prevent aggregation by sequestering with a foot in both the “foldase” and “holdase” worlds and pro- unfolded substrates or folding intermediates, i.e., by acting as a vides interesting insights into both of them. holdase. Foldase chaperones such as Hsp70, Hsp90, and GroEL Since Spy can bind to native proteins with low affinity and exhibit much more complex chaperone cycles that involve co- non-native proteins with higher affinity, the question arises as to chaperone binding and ATP hydrolysis3. Hsp70 chaperones, for why Spy does not interfere with protein function in the cell and instance, depend on J-domain proteins and nucleotide exchange why Spy does not become clogged by high-affinity interactions factors for substrate binding and release. These cochaperones with folding intermediates. We have previously kinetically and modulate Hsp70 function by regulating its ATPase cycle42. Unlike thermodynamically characterized the mechanism whereby Spy these complex and highly evolved foldase chaperones, Spy lacks binds to, folds and releases it best characterized substrate Im7 any known cochaperones or cofactors that can act to modulate (ref. 8). Periplasmic proteins are highly stable under normal

NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8 conditions, which strongly decreases the abundance of peri- molar extinction coefficients of 34,100 M−1 cm−1 for AnFld (WT, F98N, and 2A − − plasmic folding intermediates precluding the need for Spy43. mutants) and 29,000 M 1 cm 1 for AzoFld46,47. However, Spy is overproduced up to 500-fold in response to treatment by protein unfolding agents that lead to the accumu- Analytical ultracentrifugation. Sedimentation was monitored at 50 krpm by lation of (un)folding intermediates in the periplasm. Spy can absorbance at 280 nm for AnFld WT and AnFld 2A mutant. For the binding 44 experiments containing Spy and AnFld or AzoFld, sedimentation was performed at make up to 25% of the periplasmic protein content . This near 48,000 r.p.m. (180,311 × g) and monitored by absorbance at 280 nm. Sedimentation stoichiometric abundance of Spy may ensure that enough Spy velocity AUC (SV-AUC) was carried out using 450 μl loaded into two-sector epon molecules are available to handle a high client load. centerpieces with 1.2 cm path length in an An60Ti rotor in a Beckman Optima Xl-I Taken together, this work on Spy and apoflavodoxin and our analytical ultracentrifuge. Measurements were completed in intensity mode. All SV-AUC data were analyzed using UltraScan 4 software, version 4.0 and fitting past work on the effect of Spy on the folding of Im7 and Fyn SH3 procedures were completed on XSEDE clusters at the Texas Advanced Computing establish the central role of the folding landscape of the substrate Center (Lonestar, Stampede) through the UltraScan Science Gateway (https://www. in determining chaperone mechanisms. Spy can allow simple xsede.org/web/guest/gateways-listing)48. The partial specific volume (νbar) of the substrates like Im7 and SH3 to fold while bound because unfol- protein samples (Spy, AnFld WT, AnFld 2A mutant, AzoFld) was estimated within 49 ded conformations can transition to the native state on the cradle UltraScan III based on the protein sequence . Raw intensity data were converted to pseudo-absorbance by using the intensity of the air above the meniscus as a of Spy. Relatively large and topologically complex substrates like reference, and then edited. Next, two-dimensional sedimentation spectrum analysis apoflavodoxin, however, will be energetically stabilized in a (2DSA) was performed to subtract time-invariant noise, and the meniscus was fit chaperone-bound unfolded conformation. Our results therefore using ten points in a 0.05 cm range50. The arrays were fit using an S range of 1–8, – highlight the importance of substrate-specific mechanisms rather an f/f0 range of 1 4 with 64 grid points for each, 10 uniform grid repetitions, and 400 simulation points. 2DSA was then repeated at the determined meniscus to fit than the paradigmatic roles of ATP-independent chaperones in radially invariant and time-invariant noise together using ten iterations. The 2DSA protein quality control. Although the chaperone field has pri- analysis was refined by a genetic algorithm, which helps to define the solutes and to marily focused on unified mechanisms that hold true for all eliminate any false-positive solutions. The distribution between AnFld native state substrates, most experimental data for major chaperone families and AnFld intermediate IT was determined by integrating the peak intensities after performing the genetic algorithm analysis. The results from the genetic algorithm like the Hsp70s and Hsp60s suggest more substrate- and were evaluated using a Monte Carlo algorithm51. conformation-specific mechanisms4. Intuitively, multi- functionality may be advantageous for cellular and organismal fi Isothermal titration calorimetry. Thermodynamic analysis of Spy binding to tness as it enables chaperones to bind to different conforma- AnFld WT and AnFld F98N was performed by ITC experiments using a MicroCal tional states of substrates. iTC200 (Malvern Instruments). The proteins were buffer exchanged into assay buffer containing 40 mM HEPES-KOH (pH 7.5), 100 mM NaCl, using a PD10 desalting column. For the WT and F98N mutants, 550 μM Spy dimer was added in the titration syringe and 50 μM AnFld in the cell. For the 2A mutant, 1.2 mM Spy Methods dimer was added in the titration syringe and 100 μM AnFld in the cell. The fitting Protein expression and purification. The genes for AnFld and AzoFld C69A were of thermograms to one-site or two-site models was done using OriginLab that was codon-optimized for expression in E. coli and synthesized by GenScript. The fla- provided with the instrument. vodoxin genes were inserted into a pET28b-based vector with an N-terminal His6- SUMO tag by standard restriction digestion and ligation-based cloning using the fl fl fl enzymes BamHI and XhoI. The F98N, L105A, and I109A mutations were inserted Stopped- ow uorescence experiments. All stopped- ow experiments were fl into AnFld using site-directed mutagenesis with a QuickChange kit (Agilent). The performed in a KinTek SF-300X stopped- ow instrument at 25 °C. The trypto- fl fl primers used for mutagenesis are listed in Table S3. Plasmid with the spy gene was phans of both the avodoxins were excited at 295 nm and the emitted uorescence fi obtained from a lab collection8. The modified pET28b plasmid containing the signal was collected using a 320 nm long-pass lter provided with the instrument. flavodoxin or spy gene was transformed into E. coli BL21 (DE3) cells for protein Urea dependence of AnFld refolding was monitored by 11.5-fold dilution of 1.04 μ expression. Cells were grown at 37 °C to early log phase in PEM media containing M AnFld that was denatured in HN buffer containing 5 M urea into HN buffer 100 µg/ml kanamycin and then shifted to 20 °C for induction by addition of 0.1 containing various urea concentrations in a 1:10.5 mix. AnFld unfolding was mM IPTG. After 16 h of protein expression, cells were pelleted by centrifugation monitored in a similar way, except that native AnFld in HN buffer was diluted for 20 min at 5000 × g at 4 °C. The pellet was resuspended in lysis buffer containing 11.5-fold in HN buffer containing increasing concentrations of urea between 0 and fi μ 50 mM sodium phosphate, 400 mM NaCl, 10% glycerol, pH 8.0 and 0.05 μg/ml 5 M. The nal concentration of AnFld in all cases was 0.09 M. Ten to 12 traces of DNase and protease inhibitor cocktail (complete mini EDTA-free; Roche). Fol- 10 s were acquired for each data point and averaged. For refolding experiments at – fi lowing sonication for 8 min on ice, the cell lysate was centrifuged twice at 36,000 × 0.43 1.9 M nal urea concentration, two kinetic phases were observed. However, fi fi g for 30 min at 4 °C to pellet the cell debris. The supernatant was loaded along with one exponential was suf cient to t the traces from experiments at higher urea – fi 20 mM imidazole onto a Ni-His Trap column that had been equilibrated with concentrations. Average traces from unfolding experiments at 2.05 3.38 M nal fi water and lysis buffer. The column was washed with lysis buffer containing 30 mM urea concentration were t to a single kinetic phase, whereas data from experi- fi fi imidazole. His -tagged protein was eluted by adding lysis buffer containing 500 ments at higher nal urea concentrations were t to two phases. AnFld refolding in 6 μ mM imidazole to the column. Five hundred micrograms of ULP1 protease was the presence of Spy was monitored by mixing in a 1:10.5 mix, 1.04 M AnFld in added to the elution to cleave the His -SUMO tag from the protein. The elution HN buffer containing 5 M urea with HN buffer containing various Spy con- 6 – μ was then dialyzed against buffer containing 40 mM Tris, 400 mM NaCl, pH 8.0 at centrations (0 27.4 M Spy dimer after mixing). Ten to 12 traces of 10 s duration 4 °C overnight. The cleaved His tag was removed following ULP1 digestion and were acquired for each data point and averaged. AzoFld refolding in the presence of fl dialysis by passing through a Ni-His Trap column that had been equilibrated with Spy was monitored using tryptophan uorescence intensity as a signal in a similar μ lysis buffer containing 30 mM imidazole. The flow through was collected and 25 way as for AnFld, i.e., by 11.5-fold dilution of 1.5 M AzoFld in 10 mM potassium mM Tris pH 8.0 buffer was added to dilute the sample fivefold. The sample was phosphate pH 6.0 (KP) buffer containing 6 M urea into KP buffer containing – μ then loaded onto a HiTrap Q column and the protein was eluted with a NaCl various concentrations of Spy (0 182.6 M Spy dimer after mixing). Each kinetic – gradient in 25 mM Tris pH 8.0. In case of flavodoxin, the fractions containing trace shown is an average of 9 10 traces that were collected for 60 s. For stopped- fl μ holoprotein were yellow in color and were discarded. The colorless fractions ow experiments to monitor the interaction of native AnFld and Spy, 1.04 M contain purified apoflavodoxin and hence were collected and concentrated. The AnFld in HN buffer was diluted 11.5-fold into HN buffer containing various – protein was then passed through a HiLoad Superdex 75 column equilibrated with concentrations of Spy. Each kinetic trace was an average of 5 6 traces of 50 s 40 mM HEPES-KOH (pH 7.5), 100 mM NaCl (HN buffer) for desalting and buffer duration that were acquired with the auto-shutter switched on to minimize pho- fi exchange. The purified protein was then aliquoted, flash frozen in liquid nitrogen, tobleaching during the experiment. The average trace for each data point was tto and stored at −80 °C. Expression and purification of AnFld F98N, AnFld 2A, and a single exponential. In all the experiments with Spy, the background signal of Spy AzoFld were done with the same protocol. 15N-labeled AnFld WT and 2A mutant due to its tyrosine residues was removed. This was done by collecting shots of each were expressed in M9 minimal media containing 15NH Cl as the sole nitrogen concentration of Spy diluted into buffer and subtracting the average signal of Spy 4 fl source. The protocol of Marley et al.45 was used to express [U- 2H, 15N, 13C]- from the kinetic trace of avodoxin folding at that Spy concentration. Individual 2 traces were fitted to sums of exponentials in KaleidaGraph (Synergy Software) to labeled AnFld WT in M9 minimal media prepared in H2O and supplemented 15 13 obtain observed rate constants. with NH4Cl, C-glucose, and ISOGRO growth supplement (Sigma). The iso- topically labeled proteins were purified with the same protocol as that used for unlabeled proteins except that the fractions containing the holoprotein were also Fluorescence spectroscopy experiments. All fluorescence spectra were acquired fl collected. Apo avodoxin was obtained by trichloroacetic acid precipitation to on a Cary Eclipse Fluorescence Spectrophotometer. In tryptophan fluorescence- 12 remove the bound FMN . Protein concentrations were determined using the based titrations, binding was monitored by tryptophan fluorescence of

10 NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8 ARTICLE apoflavodoxin; 17.2 µM of AnFldWT or 10 µM AnFld2A in 1 ml of HN buffer was formed by mixing Spy dimer and AnFld at 1 μM final concentration in 20 mM titrated with increasing concentrations of Spy in a 10 mm quartz cuvette containing ammonium acetate solution (pH 7.4) and co-incubating on ice for ~5 min. Aqu- a magnetic stirrer. After each addition, the contents of the cuvette were manually eous samples were transferred into the gas phase via nanoelectrospray ionization mixed with a pipette and allowed to equilibrate inside the spectrophotometer for (nESI) using a capillary voltage of 900–1000 V. The ions were analyzed on a 2–3 min. Spectra were collected at 25 °C with an excitation wavelength of 295 nm modified Agilent 6560 drift tube ion mobility quadrupole time-of-flight mass 55,56 and emission wavelengths scanning from 310 to 400 nm. Both excitation and spectrometer (DTIM-Q-TOF) (Agilent) . The drying gas (N2) temperature was emission slit widths were set to 5 nm. For each addition of Spy, spectra were set to 25 °C, and the flow was reduced to 1.5 l/min. The instrument was operated collected in triplicate over a period of 0.5 min and then averaged. Average spectra with 99.9999% nitrogen gas, a front funnel pressure of 4.52 torr, a trap funnel for each titration point were corrected for dilution from the titrant additions. For pressure of 3.80 torr, a drift tube pressure of 3.947 torr, a quadrupole pressure of AnFldWT, the fluorescence emission at 340 nm for every titration point was 2.48 × 10−5 torr, and a TOF flight tube pressure of 1.699 × 10−7 torr. The IM normalized with respect to the emission of flavodoxin alone, plotted as a function separation was carried out under low field conditions (~18.5 V/cm), and the ion of Spy dimer concentration, and the binding isotherm fitted to a square hyperbola arrival time distributions were fit to Gaussian functions using CIUSuite257. The equation, Eq. (1). For AnFld2A, the fluorescence emission at 340 nm was plotted as centroids of the fit Guassian functions were converted to rotationally averaged fi DT fi a function of Spy dimer concentration and tted to a quadratic equation for tight collision cross section ( CCSN2) by a single- eld calibration using Agilent Tune 58 DT binding, Eq. (2). Mix ions as previously described . The CCSN2 measurements reported are the ðÞÀ = ¼ ðÞ*½Š=ð þ ½ŠÞ ðÞ average of three technical replicate measurements, and the error reported is the F Fi Fi Bmax L KD L 1 standard deviation. Raw data were analyzed using Agilent IM-MS Browser 10.0,  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi and mMass59,60. ¼ : à þ½ Šþ À ðÞþ½ Šþ 2À à ý Š þ ð Þ F 0 5 F0 C L KD C L KD 4 C L I 2 Reporting summary. Further information on research design is available in the Nature fl fl where F is the uorescence signal, Fi is the uorescence of AnFldWT in the absence Research Reporting Summary linked to this article. fl of Spy, [L] is the Spy dimer concentration, Bmax is the maximum uorescence fl change, KD is the dissociation constant, F0 is a uorescence correction factor, C is Data availability the concentration of AnFld2A, and I is the y-intercept. The fluorescence emission spectra of native AnFldWT and the AnFld2A mutant The raw data generated from analytical ultracentrifugation, isothermal titration were recorded by dissolving the protein in HN buffer, and the spectra of denatured calorimetry and NMR spectroscopy experiments have been included as Supplementary – AnFldWT was recorded in HN buffer containing 5 M urea. All the proteins were used Datasets 1 3 and also deposited at The Open Science Framework (OSF) (https://doi.org/ fi at a concentration of 10 µM. The spectra were collected at 25 °C with an excitation 10.17605/OSF.IO/ZPM75). The IM-MS data are available from gshare repository with fi fi wavelength of 295 nm and emission wavelengths from 310 to 400 nm. Both the slit identi er 10.6084/m9. gshare.13530872. Source data are provided with this paper. widths were 5 nm. Each spectrum was acquired three times and averaged. fl ANS binding assays were performed by monitoring the uorescence emission Received: 7 September 2020; Accepted: 8 January 2021; spectra of 1-anilino-8-napthalene sulfonate (ANS) in HN buffer in the presence and absence of 5 µM AnFld WT or 5 µM 2A mutant. The final concentration of ANS was 250 µM. ANS was excited with a wavelength of 385 nm and emission was recorded from 400 to 700 nm. The excitation and slit widths were 5 nm. The spectra shown are each an average of five scans. For chemical denaturation by urea, 10 µM AnFldWT or 5 µM AnFld2A was added to 1 ml HN buffer containing various concentrations of urea ranging from 0 to 5 References M and incubated for ~30 min at room temperature. Fluorescence spectra were 1. Bollen, Y. J. M. & Van Mierlo, C. P. M. Protein topology affects the recorded at 25 °C with excitation wavelength at 280 nm and emission monitored at appearance of intermediates during the folding of proteins with a flavodoxin- 340 nm. like fold. Biophys. Chem. 114, 181–189 (2005). 2. Chakraborty, K. et al. Chaperonin-catalyzed rescue of kinetically trapped – Circular dichroism spectroscopy experiments. Far-UV CD spectra were states in protein folding. Cell 142, 112 122 (2010). 3. Hartl, F. U., Bracher, A. & Hayer-Hartl, M. Molecular chaperones in protein acquired on a Jasco J-1500 CD spectrometer. The spectra were recorded in a quartz – cell with a path length of 0.1 cm using ~0.2 mg/ml of protein concentration. The folding and proteostasis. Nature 475, 324 332 (2011). spectra were acquired from 260 to 190 nm with a 0.5 nm data interval, 50 nm/min 4. Koldewey, P., Horowitz, S. & Bardwell, J. C. A. Chaperone-client interactions: fi – scan speed, and at 25 °C unless otherwise specified. Five scans were collected and non-speci city engenders multifunctionality. J. Biol. Chem. 292, 12010 12017 averaged. The mean residue ellipticity (MRE) was calculated using Eq. (3) (2017). 5. Mashaghi, A. et al. Reshaping of the conformational search of a protein by the MRE ¼ðMRW à ΘÞ=ð10 à d à cÞð3Þ chaperone trigger factor. Nature 500,98–101 (2013). Θ 6. Huang, C., Rossi, P., Saio, T. & Kalodimos, C. G. Structural basis for the where MRW is the mean residue weight, is the observed ellipticity (degrees), d is – the path length (0.1 cm), and c is the protein concentration in mg/ml. The average antifolding activity of a molecular chaperone. Nature 537, 202 206 (2016). CD spectra were deconvoluted with the BestSel algorithm to determine the sec- 7. Ungelenk, S. et al. Small heat shock proteins sequester misfolding proteins in fi ondary structure of the proteins in terms of the percentages of α-helix, β-sheet, near-native conformation for cellular protection and ef cient refolding. Nat. turn, and unordered components52. Commun. 7, 13673 (2016). 8. Stull, F., Koldewey, P., Humes, J. R., Radford, S. E. & Bardwell, J. C. A. Substrate protein folds while it is bound to the ATP-independent chaperone 1 15 – Nuclear magnetic resonance spectroscopy.The[ H- N] TROSY HSQC NMR Spy. Nat. Struct. Mol. Biol. 23,53–58 (2016). spectra of AnFld WT and the 2A mutant were recorded using Watergate solvent 1 9. Wu, K., Stull, F., Lee, C. & Bardwell, J. C. A. Protein folding while chaperone suppression at 25 °C on a Bruker 600 MHz spectrometer equipped with the ( H/ bound is dependent on weak interactions. Nat. Commun. 10, 4833 (2019). 19 15 F)-X broadband CryoProbe Prodigy. [U- N]-labeled proteins were exchanged 10. Horowitz, S., Koldewey, P., Stull, F. & Bardwell, J. C. Folding while bound to into NMR sample buffer (50 mM potassium phosphate buffer (pH 7.5) con- chaperones. Curr. Opin. Struct. Biol. 48,1–5 (2018). taining 0.1 M NaCl) using a PD10 desalting column (GE Healthcare). The 11. Fernández-Recio, J., Genzor, C. G. & Sancho, J. Apoflavodoxin folding samples were 0.2 mM of protein dissolved in a mixture of 90% (v/v) NMR α β 1 2 mechanism: an / protein with an essentially off-pathway intermediate. sample buffer (in H2O) and 10% (v/v) H2O. The spectra were acquired using a Biochemistry 40, 15234–15245 (2001). total of 2048 complex points in t2 and 512 increments in t1with8scansper 12. Bollen, Y. J. M., Sánchez, I. E. & Van Mierlo, C. P. M. Formation of on- and increment over a spectral width of 9.6 and 2.1 kHz in the 1Hand15Ndimen- off-pathway intermediates in the folding kinetics of Azotobacter vinelandii sions, respectively. To study the interaction of AnFld and Spy, [U-2H, 15N, 13C]- apoflavodoxin. Biochemistry 43, 10475–10489 (2004). labeled AnFld and Spy were exchanged into 50 mM potassium phosphate buffer 13. Houwman, J. A. & van Mierlo, C. P. M. Folding of proteins with a flavodoxin- (pH 7.5) without any salt using a PD10 desalting column (GE Healthcare). 15N- like architecture. FEBS J. 284, 3145–3167 (2017). TROSY spectra were collected at 25 °C with 0.2 mM AnFld dissolved in 90% 14. Kathuria, S. V., Day, I. J., Wallace, L. A. & Matthews, C. R. Kinetic traps in the (v/v) buffer (in 1H O) and 10% (v/v) D O and then titrations were done by 2 2 folding of βα-repeat Proteins: CheY initially misfolds before accessing the addition of appropriate volume of Spy to the NMR tube. A total of 2048 data – points in t2 dimension and 256 increments in t1with16scansperincrement native conformation. J. Mol. Biol. 382, 467 484 (2008). were acquired. The spectra were processed using NMRPipe suite, analyzed, and 15. Mattoo, R. U. H. & Goloubinoff, P. Molecular chaperones are nanomachines plotted in NMRFAM Sparky53,54. that catalytically unfold misfolded and alternatively folded proteins. Cell. Mol. Life Sci. 71, 3311–3325 (2014). 16. Irún, M. P., Garcia-Mira, M. M., Sanchez-Ruiz, J. M. & Sancho, J. Native Native ion mobility-mass spectroscopy. Prior to analysis, Spy and AnFld were hydrogen bonds in a molten globule: the apoflavodoxin thermal intermediate. exchanged from storage buffer into 20 mM ammonium acetate (pH 7.4) by Micro J. Mol. Biol. 306, 877–888 (2001). Bio-spin size exclusion spin columns (Bio-Rad). Spy-AnFld complexes were

NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8

17. Ayuso-Tejedor, S. et al. Design and structure of an equilibrium protein folding 47. Van Mierlo, C. P. M., Van Dongen, W. M. A. M., Vergeldt, F., Van Berkel, W. intermediate: a hint into dynamical regions of proteins. J. Mol. Biol. 400, J. H. & Steensma, E. The equilibrium unfolding of Azotobacter vinelandii 922–934 (2010). apoflavodoxin II occurs via a relatively stable folding intermediate. Protein Sci. 18. Bueno, M., Ayuso-Tejedor, S. & Sancho, J. Do proteins with similar folds have 7, 2331–2344 (1998). similar transition state structures? A diffuse transition state of the 169 residue 48. Demeler, B. & Gorbet, G. E. in Analytical Ultracentrifugation: apoflavodoxin. J. Mol. Biol. 359, 813–824 (2006). Instrumentation, Software, and Applications (eds. Uchiyama S. et al.) 119–143 19. Konrat, R. NMR contributions to structural dynamics studies of intrinsically (Springer, 2016). disordered proteins. J. Magn. Reson. 241,74–85 (2014). 49. Demeler, B., Brookes, E. & Nagel-Steger, L. in Methods in Enzymology Vol. 20. Hyung, S. J. & Ruotolo, B. T. Integrating mass spectrometry of intact protein 454 (eds. Abelson, J. et al.) Ch. 4, 87–113 (Academic Press, 2009). complexes into structural proteomics. Proteomics 12, 1547–1564 (2012). 50. Brookes, E., Cao, W. & Demeler, B. A two-dimensional spectrum analysis for 21. Mason, E. A. & McDaniel, E. W. Transport Properties of Ions in Gases (Wiley, sedimentation velocity experiments of mixtures with heterogeneity in 1988). molecular weight and shape. Eur. Biophys. J. 39, 405–414 (2010). 22. Gadkari, V. V. et al. Investigation of sliding DNA clamp dynamics by single- 51. Brookes, E. & Demeler, B. Genetic algorithm optimization for obtaining molecule fluorescence, mass spectrometry and structure-based modeling. accurate molecular weight distributions from sedimentation velocity Nucleic Acids Res. 46, 3103–3118 (2018). experiments. Prog. Colloid Polym. Sci. 131,33–40 (2006). 23. Bush, M. F. et al. Collision cross sections of proteins and their complexes: a 52. Micsonai, A. et al. BeStSel: a web server for accurate protein secondary calibration framework and database for gas-phase structural biology. Anal. structure prediction and fold recognition from the circular dichroism spectra. Chem. 82, 9557–9565 (2010). Nucleic Acids Res. 46, W315–W322 (2018). 24. Maldonado, S. et al. Salt-induced stabilization of apoflavodoxin at neutral pH 53. Delaglio, F. et al. NMRPipe: a multidimensional spectral processing system is mediated through cation-specific effects. Protein Sci. 11, 1260–1273 (2002). based on UNIX pipes. J. Biomol. NMR 6, 277–293 (1995). 25. Horowitz, S. et al. Visualizing chaperone-assisted protein folding. Nat. Struct. 54. Lee, W., Tonelli, M. & Markley, J. L. NMRFAM-SPARKY: enhanced software Mol. Biol. 23, 691–697 (2016). for biomolecular NMR spectroscopy. Bioinformatics 31, 1325–1327 26. He, L., Sharpe, T., Mazur, A. & Hiller, S. A molecular mechanism of (2015). chaperone-client recognition. Sci. Adv. 2, e1601625 (2016). 55. Vallejo, D. D. et al. A modified drift tube ion mobility-mass spectrometer for 27. Koldewey, P., Stull, F., Horowitz, S., Martin, R. & Bardwell, J. C. A. Forces charge-multiplexed collision-induced unfolding. Anal. Chem. 91, 8137–8146 driving chaperone action. Cell 166, 369–379 (2016). (2019). 28. Langdon, G. M. et al. Anabaena apoflavodoxin hydrogen exchange: on the 56. Gadkari, V. V. et al. Enhanced collision induced unfolding and electron stable exchange core of the α/β(21345) flavodoxin-like family. Proteins Struct. capture dissociation of native-like protein ions. Anal. Chem. 92, 15489–15496 Funct. Genet. 43, 476–488 (2001). (2020). 29. Campos, L. A., Bueno, M., Lopez-Llano, J., Jiménez, M. Á. & Sancho, J. 57. Polasky, D. A., Dixit, S. M., Fantin, S. M. & Ruotolo, B. T. CIUSuite 2: Next- Structure of stable protein folding intermediates by equilibrium φ-analysis: the Generation Software for the Analysis of Gas-Phase Protein Unfolding Data. apoflavodoxin thermal intermediate. J. Mol. Biol. 344, 239–255 (2004). Analytical Chemistry 91, 3147–3155 (2019). 30. Nelson, E. D. & Grishin, N. V. Alternate pathways for folding in the 58. Stow, S. M. et al. An interlaboratory evaluation of drift tube ion mobility-mass flavodoxin fold family revealed by a nucleation-growth model. J. Mol. Biol. spectrometry collision cross section measurements. Anal. Chem. 89, 358, 646–653 (2006). 9048–9055 (2017). 31. Ptitsyn, O. B. in Advances in Protein Chemistry (eds Anfinsen, C. B. et al.) 59. Strohalm, M., Hassman, M., Košata, B. & Kodíček, M. mMass data miner: An Vol. 47, 83–229 (Elsevier, 1995) open source alternative for mass spectrometric data analysis. Rapid Commun. 32. Lindhoud, S., Pirchi, M., Westphal, A. H., Haran, G. & Van Mierlo, C. P. M. Mass Spectrom. 22, 905–908 (2008). Gradual folding of an off-pathway molten globule detected at the single- 60. Strohalm, M., Kavan, D., Novák, P., Volný, M. & Havlíček, V. MMass 3: A molecule level. J. Mol. Biol. 427, 3148–3157 (2015). cross-platform software environment for precise analysis of mass 33. Houwman, J. A., André, E., Westphal, A. H., Van Berkel, W. J. H. & Van spectrometric data. Anal. Chem. 82, 4648–4651 (2010). Mierlo, C. P. M. The ribosome restrains molten globule formation in stalled nascent flavodoxin. J. Biol. Chem. 291, 25911–25920 (2016). 34. Engel, R. et al. Macromolecular crowding compacts unfolded apoflavodoxin Acknowledgements and causes severe aggregation of the off-pathway intermediate during We thank Sheena Radford, Kevin Wu, and Frederick Stull for helpful discussions and for apoflavodoxin folding. J. Biol. Chem. 283, 27383–27394 (2008). critical reading the manuscript. We thank Ke Wan in the Bardwell laboratory for protein 35. Quan, S. et al. Super Spy variants implicate flexibility in chaperone action. purification. We are very grateful to Debashish Sahu and Minli Xing at the BioNMR Core Elife 2014,1–22 (2014). of the University of Michigan and Bikash Sahoo and Scott Gorman for their help and 36. Bornemann, T., Holtkamp, W. & Wintermeyer, W. Interplay between trigger guidance with NMR experiments. J.C.A.B. is an investigator at the Howard Hughes factor and other protein biogenesis factors on the ribosome. Nat. Commun. 5, Medical Institute, which funded this work. V.V.G. is supported by the Agilent Thought 4180 (2014). Leader Award presented to B.T.R. 37. Qu, J., Mayer, C., Behrens, S., Holst, O. & Kleinschmidt, J. H. The trimeric periplasmic chaperone Skp of Escherichia coli forms 1:1 complexes with outer Author contributions membrane proteins via hydrophobic and electrostatic interactions. J. Mol. R.M., V.V.G., B.A.M., B.T.R. and J.C.A.B. designed the study; R.M., V.V.G. and B.A.M. Biol. 374,91–105 (2007). performed the experiments; R.M., V.V.G., B.A.M., C.P.M.v.M. and J.C.A.B. analyzed the 38. Haslbeck, M., Weinkauf, S. & Buchner, J. Small heat shock proteins: simplicity data; and R.M., V.V.G., B.A.M. and J.C.A.B. wrote the manuscript. meets complexity. J. Biol. Chem. 294, 2121–2132 (2019). 39. Basha, E., O’Neill, H. & Vierling, E. Small heat shock proteins and α-crystallins: dynamic proteins with flexible functions. Trends Biochem. Sci. 37, 106–117 Competing interests (2012). The authors declare no competing interests. 40. Jovcevski, B. et al. Phosphomimics destabilize Hsp27 oligomeric assemblies – and enhance chaperone activity. Chem. Biol. 22, 186 195 (2015). Additional information 41. Shashidharamurthy, R., Koteiche, H. A., Dong, J. & Mchaourab, H. S. Supplementary information Mechanism of chaperone function in small heat shock proteins. J. Biol. Chem. The online version contains supplementary material 280, 5281–5289 (2005). available at https://doi.org/10.1038/s41467-021-21120-8. 42. Rosenzweig, R., Nillegoda, N. B., Mayer, M. P. & Bukau, B. The Hsp70 Correspondence chaperone network. Nat. Rev. Mol. Cell Biol. 20, 665–680 (2019). and requests for materials should be addressed to J.C.A.B. 43. Park, C., Zhou, S., Gilmore, J. & Marqusee, S. Energetics-based protein Peer review information profiling on a proteomic scale: identification of proteins resistant to Nature Communications thanks Paolo De Los Rios, Xavier proteolysis. J. Mol. Biol. 368, 1426–1437 (2007). Salvatella, and other, anonymous, reviewers for their contributions to the peer review of 44. Quan, S. et al. Genetic selection designed to stabilize proteins uncovers a this work. Peer review reports are available. chaperone called Spy. Nat. Struct. Mol. Biol. 18, 262–269 (2011). Reprints and permission information 45. Marley, J., Lu, M. & Bracken, C. A method for efficient isotopic labeling of is available at http://www.nature.com/reprints recombinant proteins. J. Biomol. NMR 20,71–75 (2001). fl Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 46. Genzor, C. G. et al. Conformational stability of apo avodoxin. Protein Sci. 5, fi 1376–1388 (1996). published maps and institutional af liations.

12 NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21120-8 ARTICLE

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.

© The Author(s) 2021

NATURE COMMUNICATIONS | (2021) 12:851 | https://doi.org/10.1038/s41467-021-21120-8 | www.nature.com/naturecommunications 13